Income Mobility in Ecuador: New evidence from individual income tax - - PowerPoint PPT Presentation

Income Mobility in Ecuador: New evidence from individual income tax returns

Liliana CANO University of Toulouse - Lereps September 5th, 2014 INEQUALITY measurement, trends, impacts and policies

1

Outline

• Goals of the paper
• Motivation
• Literature review
• Data and methodology
• Main findings
• Concluding remarks

2

• 1. Goals of this paper
• Analyze income mobility in Ecuador with a focus on the top

and on the middle of the distribution. This study is based on income tax returns database from 2004 to 2011.

• We study whether the evolution of top income shares has

been accompanied by an increase or a decrease in mobility for the high income groups.

• We study whether there is a surge of an Ecuadorian middle

class.

• We analize the factors associated with income mobility over

the 2008-2011 period.

3

• 2. Motivations
• The first motivation is based on the growing interest

in the study of income inequality at the top of the distribution using income tax data and national accounts (Piketty 2001, 2003).

• Method: Kuznets (1953), Atkinson and Piketty (2007,

2010).

• Top income series in more than 26 countries.

4

5

• 2. Motivations
• The second motivation is based on the study of

intragenerational mobility.

• A recent economic report from the World Bank

documented that almost 43% of Latin American individuals had experienced changes in their economic status over the last years.

• Mostly upward movements.
• In Ecuador estimates of income mobility are

scarce mainly due to lack of appropriate data.

6

• 3. Literature review
• Sociological and economic approaches of mobility. In this paper we focus on

an economic approach.

• Literature on income mobility is vast : there is not a harmonized framework of

mobility measurement. Mobility might connotes different ideas to different researchers.

• An important review of conceptual and methodological issues is provided in

Fields (2000), Atkinson et al (2001), Jenkins and Van Kerm (2006), ,Fields (2008), Burkhauser and Couch (2011), Jantti and Jenkins (2013)

7

• Drawing on the taxonomy by Fields (2001) :

– Two different magnitudes : intra-generational and intergenerational – Three broad conceptions of mobility – These concepts do capture very different aspects of mobility

Directional Non - directional

Income Shares Positions (rank)

• Mobility as movement
• Mobility as time independence
• Mobility as Equalizer of long -

term

IM (D) IM (ND) SM PM MTI ELTI

8

• 3. Literature review on top incomes

Author Country Data Findings

Intra-generational mobility

Auten and Gee (2009) Auten et al. (2013) United States : 1987 – 2005 United States : 2005 - 2010 Income tax returns 40% placed in the top 1% remains at the top in

• 2005. And more than

50% moved to a different centile. Kopczuk (2010) United States, since 1937 Social Security administration There is not mobility at the top. 60% probability

• f remains at the top.

Saez & Veall (2005) Canada : 1982 - 2000 Income tax returns

Not mobility at the top ; probability stay 60%

Landais (2009) France : 1996 - 2006 Income tax returns

Not mobility at the top : probability stay 67%

Intergenerational mobility

Chetty (2014) United States : 1996 - 2012 Federal income tax

Mobility depends on the geographical area and the fact of moving is driving by factors like ethnic origin, parent’s income level, family characteristics, social networks, etc. But not for top 1%.

Bjorklund et al (2012) Sweden Income tax returns Transmitions between fathers and sons at the top is very strong. Elasticity of almost 0,9.

9

3.1 Hypotheses

• H1: Income inequality declining trend has not improve income

mobility at the very top.

• H2: There is a high degree of upward income mobility in

Ecuador over the past years.

• H3: Upward mobility is mainly explained by the initial position

in the income distribution.

• H4: The upward economic effect of education on income

mobility should be more or as important as initial position.

10

• Longitudinal micro data from income tax returns

from 2004 to 2011. The universe of tax filers.

• Information from 3 different types of tax form:

– 107: salaries and wages – 102a: wages, self-employment income, capital returns and

• ther possible source of income.

– 102: income information (labor and capital) for individuals who required to keep accounting books.

• For instance: 2.3 million tax filers in 2011
• Unit of observation : individuals
• Anonymous data
• 4. Data

11

• Advantages of tax statistics:
• Tax data are relatively homogenous within a country.
• Provide a better picture of the top and the middle of the distribution.
• Provide composition of incomes.
• Real panel database.
• Disadvantages of tax statistics:
• Evasion and elusion.
• Tax reforms change the definition of income across time.
• 4. Data

12

• Information on individual characteristics of tax filers

from the Ecuadorian Civil Registry.

• Six explanatories variables.

– Initial position in the income distribution: i.e. 10 deciles – Age: -20 years, 20-29, 30-39, 40-49, 50-59 and 60 + years – Gender: 1=men, 0=women – Marital status: 1=married, 0=otherwise – Level of education: 1=high school and more, 0 less than high school. – Geographical region: North, Center, South, Coast, Pichincha and Guayas.

• 4. Data

13

• 5. Methodology

1. We construct annual series on top shares of income by relating the amounts of individual income tax returns (numerator of the share) to a comparable control total for full population (denominator of the share).

– Income definition : income reported on tax returns that includes salaries and wages, self-employment and small business, rents and capital income (interest and dividends) and items reported as

• ther income: long term capital gains, inheritances, donations and

legal deductions to obtain income. – Income definition is before personal income taxes and employee payroll taxes. – Top 1% (P99 – P100), top 0.5% (P99.5-100), top 0.1% (P99.9-100), top 0,01 (P99.99-100) etc.

14

• 5. Methodology

– To construct incomes shares : income of each fractile / control income reported by household surveys. – Period: 2004-2011

• Control for total income

– Total income from Ecuadorian household survey ENEMDU – wages, self-employment, capital, transfers, secondary

• income. (~65% of GDP)
• Control for total population

– Information from ENEMDU – Adult population (economically active population) age 20 and older.

15

16

• 5. Methodology
• Top income persistence: we calculate the probability
• f remaining in the top 1%, top 0.1% and the top

0.01% after different periods of time (Saez and Veall,

2005; Landais, 2009)

• Transitions between top fractiles: Using transitions

matrices we examine movements of individuals across top fractiles.

• 2. We analyze mobility for all tax filers from 2004-

2011.

• Income deciles are constructed relative to the tax filing

population.

• 5. Methodology

17

3. Factors associated with mobility in Ecuador 2008 - 2011: we estimate transition probabilities of upward or downward movements while controlling for control variables: – Counting procedure – Multinomial logit model – Generalized ordered logit model

• Tax filers in 2008: 1.9 million
• Tax filers in 2011: 2.3 million

– Control by initial position: 1.4 million of observations – With all control variables: 737.891 observations

Methodological limitation : how many income centiles? We use two additional methods:

– A multinomial logit model to assess upward or downward movements of at least 10 centiles from a given initial position. – A logistic model where the dependent variable measures the change in the percentile position of an individual from 2008 to 2011.

18

Main findings

19

20 (1) (2) (3) (4) (5) (6) (7) Full Population 9 408 267 $9 417 $17 896 P90 $7 141 $13 572 Top 10-5% 470 413 $28 648 $54 446 P95 $12 898 $24 512 Top 5-1% 376 331 $32 350 $61 481 P99 $33 800 $64 236 Top 1-0.5% 47 041 $91 712 $174 298 P99.5 $47 537 $90 342 Top 0.5-0.1% 37 633 $102 172 $194 176 P99.9 $98 236 $186 695 Top 0.1-0.05% 4 704 $299 473 $569 145 P99.95 $138 201 $262 648 Top 0.05-0.01% 3 763 $337 840 $642 059 P99.99 $313 641 $596 071 Top 0.01% - Top 0,001% 847 $773 507 $1 470 039 P99.999 $1 132 662 $2 152 608 Top 0,001% 94 $2 893 022 $5 498 146

Note : In 2011 for Ecuador PPP US$ 1 = 0,52618 Note 2 : Computations are based on income tax returns statistics.

Table 4. Thresholds and average incomes in top groups within the top percentile, Ecuador 2011

Thresholds Income threshold Income Groups Number of tax units Average income US$ US$ (PPP) US$ US$ (PPP)

21

5 10 15 20 25 30 35 40 2004 2005 2006 2007 2008 2009 2010 2011

Income share (%)

Source: Author's calculation based on individual income tax returns. Number of tax units is estimated. Total income is estimated from household surveys. Top shares are obtained from income tax returns statistics.

Fig 1. Income Share of the top 1 percent in Ecuador 2004 - 2011

Top 1%

In 2011 almost 20% of total income goes to the top 1% of the population

22

2 4 6 8 10 12 2004 2005 2006 2007 2008 2009 2010 2011

Income share (%)

Source: Author's calculation based on individual income tax returns. Number of tax units is estimated. Total income is estimated from household surveys. Top shares are obtained from income tax returns statistics.

Fig 2. Top 1 - 0.5%, Top 0.5 - 0.1%, Top 0.1%

Ecuador, 2004 - 2011

Top 1-0.5% Top 0.5-0.1% Top 0.1%

23

0% 10% 20% 30% 40% 50% 60% 70% 80% 90%

2004 2005 2006 2007 2008 2009 2010 Probability of staying in top grup

Fig 3. Evolution of top income mobility in Ecuador (2004 - 2011) Income mobility among the P99 - P100

• A. Probability of staying in the top 1% group

1 year after 2 year after 3 year after

Probabilities on average : 65%, 56%, 49%

24

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 2004 2005 2006 2007 2008 2009 2010 Probability of staying in top grup

Source : Author's computations using individual income tax returns

Fig 5. Evolution of top income mobility in Ecuador (2004 - 2011) Income mobility among the P99.99 - P100

• C. Probability of staying in the top 0,01% group

1 year after 2 year after 3 year after

Probabilities on average : 32%, 19%, 15%

25

Origin 2004

Bottom 95% Top 5% Top 1% Top 0,5% Top 0,1% Top 0,05% Top 0,01% Total Bottom 95% 77,4 17,4 2,4 2,2 0,3 0,3 0,1 100 Top 5% 44,3 48,9 4,1 2,4 0,2 0,1 0,0 100 Top 1% 19,8 50,0 17,7 10,9 0,9 0,6 0,1 100 Top 0,5% 19,4 29,3 21,5 25,1 2,8 1,7 0,3 100 Top 0,1% 23,9 18,9 10,2 30,3 9,3 6,4 1,1 100 Top 0,05% 24,0 17,2 9,9 23,6 10,6 11,2 3,6 100 Top 0,01% 35,0 17,1 7,4 12,5 5,8 12,1 10,1 100 Total 61,7 29,7 4,3 3,5 0,5 0,4 0,1 100

(a) Top series are obtained from income tax returns statistics (b)For top shares, control population and control total income are estimated from household surveys

Table 5 : Top Income Mobility in Ecuador (a,b) Transitions between income fractiles 2004 - 2011 % of net fractile members Destination 2011

• Diagonal entries present the « stayers groups»
• Rows correspond to top percentiles at origin (2004)
• Columns correspond to top percentiles at destination (2011)

Top 1%:

• 82% (100%-17%)moved by 2011
• 13% moved up and 70% moved down.
• 50% moved down to top 5%

Top 0.1% = 7.5% moved up 83% moved down but only 23.9% had dropped to the bottom 95%

26

Origin 2008

Bottom 95% Top 5% Top 1% Top 0,5% Top 0,1% Top 0,05% Top 0,01% Total Bottom 95% 86,7 12,0 0,7 0,5 0,1 0,0 0,0 100 Top 5% 24,1 65,2 7,1 3,3 0,2 0,1 0,0 100 Top 1% 16,2 30,3 29,3 22,2 1,3 0,7 0,1 100 Top 0,5% 19,1 18,6 14,0 37,2 6,8 3,7 0,5 100 Top 0,1% 20,3 16,3 8,1 22,9 13,5 17,1 2,0 100 Top 0,05% 20,7 14,6 8,3 18,7 8,5 21,6 7,6 100 Top 0,01% 24,2 16,8 5,9 10,6 4,7 13,0 24,9 100 Total 71,0 23,2 2,9 2,3 0,3 0,2 0,1 100

(a) Top series are obtained from income tax returns statistics (b) For top shares, control population and control total income are estimated from household surveys

Table 6 : Top Income Mobility in Ecuador (a,b) Transitions between income fractiles 2008 - 2011 % of net fractile members Destination 2011

• Diagonal entries present the « stayers groups»
• Rows correspond to top percentiles at origin (2008)
• Columns correspond to top percentiles at destination (2011)

Top 1%:

• 71% (100%-29.3%)moved by 2011
• 24.3% moved up
• 30. 3% moved down to top 5%
• 16% had dropped to the bottom 95%

Top 0.1% = 87% moved. 20% dropped to bottom 95%.

Factors associated with income mobility

27

Markov transitions probabilities

1. Counting procedure:

Where:

• is the number of tax filers who were in decile in

year t-3 an now are in decile in year t

• is the probability of a tax filer being in decile in

year t, given that he was in state in year t - 3

28

=

• Σ

• 2. Multinomial logit model
• where X is the vector of explanatory variables for the ith observation and j

is the vector of parameters to be estimated for each jth outcome.

• The dependent variable takes ten different outcomes: 1 if first decile, 2 if

second decile, 3 if third decile, . . . 10 if ten decile.

29

1 1 + Σ

• exp() , if = 1

Pr = = exp() 1 + Σ

• exp() , if = 2, 3, … 10

• 3. Generalized ordered logit model
• Because of natural ordering in the deciles positions,

predicted probabilities are calculated:

• where # are ordered estimated cutpoints and where j ranges

from 1 to 10.

30

ex$(# − ) 1 + exp(#− ) , for = 1

'()(*+, -./+) 0 '()(*+, -./+) − '() *+12, -./+12 0'() *+12, -./+12 , for = 2 to 4 − 1

1 − exp #5, − 5, 1 + exp #5, − 5, , for j = J

Pr = =

31

Panel A: full population without control variables (probabilities obtained by counting transitions or predicted from generalized ordered logit model, or from multinomial logit model) Origin 2008 Destination 2011 N % DECILE 1 2 3 4 5 6 7 8 9 10 Total Total 3 90 940 6,5% 1 16,7 12,8 12,9 11,4 10,6 9,1 7,0 5,5 5,5 8,6 100,0 42,4 110 400 7,8% 2 10,6 13,0 14,8 15,9 15,7 12,5 7,8 5,2 2,8 1,8 100,0 46,4 129 258 9,2% 3 6,8 8,8 12,8 18,2 18,6 14,7 9,6 5,1 3,3 1,9 100,0 51,6 142 433 10,1% 4 4,7 6,1 9,9 24,7 20,5 14,9 9,2 5,2 2,9 1,9 100,0 60,2 151 185 10,7% 5 3,5 4,3 6,1 10,2 22,3 24,0 15,2 7,9 4,0 2,4 100,0 61,5 156 316 11,1% 6 2,4 2,7 3,7 3,8 8,4 26,9 27,8 14,5 6,2 3,7 100,0 69,2 160 197 11,4% 7 1,7 1,6 2,2 1,9 2,9 7,6 29,2 31,7 16,0 5,1 100,0 76,9 162 898 11,6% 8 1,4 1,1 1,5 1,2 1,5 2,8 7,9 34,1 38,6 9,7 100,0 82,5 155 070 11,0% 9 1,8 1,2 1,5 1,3 1,5 2,2 4,1 10,6 42,1 33,7 100,0 86,4 149 800 10,6% 10 2,9 1,1 1,8 1,4 1,4 1,8 2,7 4,3 11,7 71,0 100,0 87,0 1 408 497 100,0%

• A high degree of mobility especially tax filers of 2nd and 3rd decile : 87% moved by 2011.
• Between 75% and 66% of people placed in the 4th and the 8th decile.
• A much larger portion of individuals moved up to a higher decile than dropped to a lower

decile.

• Diagonal entries increases with higher deciles.

32

Panel B: sub-sample without control variables (probabilities obtained by counting transitions or predicted from multinomial logit model) or with control variables (probabilities from multinomial logit model) Origin 2008 Destination 2011 N % DECILE 1 2 3 4 5 6 7 8 9 10 Total Total 3 28 996 3,9% 1 15,1 15,0 14,4 13,4 12,6 11,1 7,2 4,6 3,4 3,3 100,0 44,4 50 954 6,9% 2 10,3 12,4 14,5 14,9 16,1 13,6 8,7 5,1 2,9 1,4 100,0 45,4 61 086 8,3% 3 6,8 8,4 11,8 16,7 19,2 15,6 11,1 5,6 3,4 1,6 100,0 51,5 68 311 9,3% 4 4,8 6,0 9,0 24,0 21,6 15,0 9,9 5,3 2,9 1,6 100,0 60,5 85 100 11,5% 5 3,2 4,1 5,9 9,4 23,1 24,6 15,9 8,3 3,7 1,9 100,0 63,5 92 512 12,5% 6 2,0 2,5 3,3 3,3 7,7 27,9 29,3 15,1 5,8 3,1 100,0 72,3 95 860 13,0% 7 1,2 1,5 1,9 1,6 2,7 6,9 30,2 36,3 13,7 4,0 100,0 80,2 95 297 12,9% 8 0,9 1,0 1,1 0,9 1,2 2,4 7,6 40,0 37,0 7,9 100,0 84,9 86 509 11,7% 9 0,9 0,8 1,1 0,8 1,0 1,7 3,3 9,5 48,3 32,6 100,0 90,4 73 266 9,9% 10 1,0 0,6 1,0 0,8 0,9 1,3 1,9 3,4 11,4 77,7 100,0 92,6 737 891 100,0%

• Results suggest that individuals placed into the middle deciles (3th to 8th) are

more likely to experience upward movements (56% on average) than a downward movement (19% on average) or simply no movement (25% on average).

33

Panel C: sub-sample with control variables (transition probabilities from generalized ordered logit model) Origin 2008 Destination 2011 N % DECILE 1 2 3 4 5 6 7 8 9 10 Total Total 3 28 996 3,9% 1 14,8 14,8 14,3 13,6 12,9 11,3 7,2 4,7 3,2 3,2 100,0 43,9 50 954 6,9% 2 10,3 12,4 14,4 14,8 16,3 13,7 8,7 5,1 2,8 1,4 100,0 45,5 61 086 8,3% 3 6,8 8,4 11,8 16,4 19,2 15,7 11,1 5,7 3,3 1,5 100,0 51,3 68 311 9,3% 4 4,8 6,0 9,1 23,5 21,6 15,2 10,0 5,4 2,8 1,6 100,0 60,3 85 100 11,5% 5 3,2 4,1 5,9 9,4 22,8 24,7 16,1 8,3 3,7 1,9 100,0 63,6 92 512 12,5% 6 1,9 2,4 3,2 3,6 7,8 27,5 29,4 15,2 5,9 3,1 100,0 72,1 95 860 13,0% 7 1,2 1,4 1,9 1,8 2,8 7,0 29,8 36,3 13,9 4,0 100,0 80,0 95 297 12,9% 8 0,8 0,9 1,1 1,0 1,3 2,5 7,7 39,7 37,0 8,0 100,0 84,7 86 509 11,7% 9 0,8 0,7 0,9 0,9 1,1 1,7 3,4 9,5 48,0 32,8 100,0 90,3 73 266 9,9% 10 0,9 0,5 0,9 0,8 0,9 1,3 2,0 3,6 11,5 77,7 100,0 92,8 737 891 100,0% This table reports mean values of transition probabilities from positions in the income distribution in 2008 to decile positions in 2011. Deciles are computed on the entire tax filing population but transitions probabilities are computed for survivors in 2011. In models with control variables, predicted probabilities are conditioned by previous position in income distribution, birth region, age, gender, marital status, and education. The most important probability by decile is in italic and in blue. The three most important probabilities are in bold. Their sum is in column “Total 3”.

Results suggest that individuals placed into the middle deciles (3th to 8th) are more likely to experience upward movements (56% on average) than a downward movement (19% on average) or simply no movement (25% on average).

34

• Changes in predicted probabilities suggest that having a scholar degree highly

influences the probability of moving across the income distribution.

• Probabilities of moving up are higher for those starting in the 6th decile and

who have a scholar degree.

• Probabilities of falling to the lowest deciles are higher for those starting in the

6th decile and without a scholar degree.

How many income centiles?

35

• 4. Strong movements predicted by a multinomial logit

model

• where X is the vector of explanatory variables for the ith observation and j is the

vector of parameters to be estimated for each jth outcome.

• The dependent variable takes three different outcomes: 1 if not movement or weak

movement, 2 if strong upward movement, and 3 if strong downward movement. Where strong means a movement superior to 10 centiles.

• Probabilities of strong upward or downward mobility are estimated relative to the

base category of « weak movement »

36

1 1 + Σ

8

exp() , if = 1 Pr = = exp() 1 + Σ

8

exp() , if = 2, 3

37

Downward and upward movements of at least 10 centiles (Logit Multinomial)

(1) (2) (3) (4) (5) (6) upward downward upward downward upward downward upward downward upward downward upward downward dec1 3.053* na 2.758* na 2.438* na 1.144* na 0.849* na 0.635* na (0.023) (0.028) (0.039) (0.031) (0.024) (0.020) dec2 2.484* 0.155* 2.247* 0.145* 2.155* 0.133* 0.997 0.135* 0.742* 0.130* 0.555* 0.113* (0.017) (0.002) (0.021) (0.003) (0.029) (0.004) (0.026) (0.004) (0.020) (0.004) (0.016) (0.004) dec3 2.182* 0.362* 1.973* 0.338* 2.067* 0.345* 0.961 0.365* 0.710* 0.352* 0.532* 0.307* (0.014) (0.004) (0.018) (0.005) (0.027) (0.007) (0.025) (0.010) (0.019) (0.010) (0.015) (0.010) dec4 0.969* 0.332* 0.877* 0.311* 0.830* 0.298* 0.394* 0.329* 0.290* 0.319* 0.217* 0.279* (0.006) (0.003) (0.008) (0.004) (0.010) (0.005) (0.010) (0.008) (0.008) (0.008) (0.006) (0.008) dec5 0.862* 0.376* 0.778* 0.352* 0.717* 0.310* 0.342* 0.350* 0.237* 0.345* 0.176* 0.299* (0.005) (0.003) (0.007) (0.004) (0.008) (0.005) (0.009) (0.009) (0.006) (0.009) (0.005) (0.009) dec6 0.673* 0.287* 0.608* 0.270* 0.537* 0.221* 0.270* 0.263* 0.172* 0.264* 0.127* 0.228* (0.004) (0.002) (0.005) (0.003) (0.006) (0.004) (0.007) (0.006) (0.004) (0.007) (0.004) (0.007) dec7 0.556* 0.217* 0.505* 0.205* 0.409* 0.156* 0.207* 0.187* 0.125* 0.191* 0.092* 0.165* (0.003) (0.002) (0.004) (0.002) (0.005) (0.003) (0.005) (0.005) (0.003) (0.005) (0.003) (0.005) dec8 0.314* 0.176* 0.285* 0.167* 0.192* 0.116* 0.099* 0.144* 0.057* 0.150* 0.042* 0.129* (0.002) (0.001) (0.003) (0.002) (0.002) (0.002) (0.003) (0.004) (0.002) (0.004) (0.001) (0.004) dec9 0.093* 0.221* 0.084* 0.209* 0.059* 0.131* 0.033* 0.176* 0.019* 0.185* 0.014* 0.160* (0.001) (0.002) (0.001) (0.002) (0.001) (0.002) (0.001) (0.004) (0.001) (0.005) (0.000) (0.005) dec10 na 0.250* na 0.234* na 0.130* na 0.181* na 0.191* na 0.166* (0.002) (0.003) (0.002) (0.004) (0.005) (0.005) pichincha 1.074* 1.121* 1.215* 1.246* 1.185* 1.210* 1.117* 1.238* 1.118* 1.238* (0.009) (0.012) (0.013) (0.018) (0.013) (0.018) (0.012) (0.018) (0.012) (0.018) guayas 1.227* 1.130* 1.474* 1.301* 1.436* 1.255* 1.351* 1.275* 1.346* 1.273* (0.010) (0.012) (0.017) (0.020) (0.017) (0.019) (0.016) (0.020) (0.016) (0.020) coast 1.030* 1.087* 1.046* 1.112* 1.066* 1.133* 1.053* 1.133* 1.045* 1.128* (0.009) (0.012) (0.012) (0.018) (0.013) (0.018) (0.013) (0.018) (0.013) (0.018) center 1.119* 0.934* 1.107* 0.897* 1.130* 0.910* 1.077* 0.923* 1.073* 0.922* (0.010) (0.011) (0.013) (0.015) (0.014) (0.015) (0.013) (0.015) (0.013) (0.015) south 1.116* 0.979 1.241* 0.988 1.291* 1.012 1.241* 1.032 1.234* 1.030 (0.011) (0.012) (0.016) (0.017) (0.017) (0.018) (0.016) (0.018) (0.016) (0.018) age19 1.602* 1.310* 1.303* 1.367* 1.348* 1.386* (0.047) (0.047) (0.039) (0.050) (0.040) (0.051) age20_29 2.555* 1.124* 2.007* 1.183* 2.061* 1.193* (0.058) (0.022) (0.046) (0.024) (0.048) (0.024) age30_39 2.173* 0.823* 1.770* 0.862* 1.812* 0.868* (0.049) (0.016) (0.041) (0.017) (0.042) (0.017)

Continued on next page

38

age40_49 1.608* 0.634* 1.338* 0.663* 1.364* 0.666* (0.037) (0.013) (0.031) (0.014) (0.032) (0.014) age50_59 1.073* 0.524* 0.965 0.540* 0.975 0.541* (0.027) (0.012) (0.024) (0.012) (0.024) (0.012) gender 1.199* 1.118* 1.667* 1.319* (0.008) (0.009) (0.027) (0.027) married 1.044* 0.964* (0.007) (0.008) education 2.015* 0.845* (0.015) (0.008) marriedman 1.090* 0.982 (0.009) (0.010) marriedwoman 0.974 0.937* (0.010) (0.013) educman 1.809* 0.802* (0.015) (0.009) educwoman 2.874* 1.009 (0.043) (0.020) Obs. 1 408 497 1 408 497 737 891 737 891 737 891 737 891 Chi2 statistic 430313.03 430980.62 268284.33 271640.32 277792.66 278645.23 Log pseudolikelihood

• 1174173.92
• 1173263.96
• 587039.69
• 581542.11
• 575765.08
• 575336.37

Exponentiated coefficients * p<0.01 na: coefficients non available because they cannot be estimated (no upward movement for dec10 and no downward movement for dec1) Omitted categories are north, age60.

• 5. Modeling centile effects
• We measure the change in the centile position from the base

period (2008) to the end of the period (2011).

• The simplest would be the difference in the two percentiles

positions : 40th centile to the 50th centile = moved up 10 centiles.

• However, this variable presents consistency problems since

the centile range is bounded by zero and 100.

39

• 5. Modeling centile effects
• We use a logit transformation of the dependent variable.
• Following Auten and Gee (2009) the dependent variable is defined as:

y = logit(decent) = ln (dcent/ (1 − decent)) decent = 1/2(endcentile − startcentile) + 50 100

• Where decent is a transformation scaled in such a manner that individuals

whose income remain the same at the end of the period, hold a dependent variable of « zero »

• This transformation allows us to use logistic regression to model

movement in the population.

40

41 Factors associated with income mobility in Ecuador Regression results, 2008 - 2011

(1) (2) (3) (4) (5) (6) dcent centile effect dcent centile effect dcent centile effect dcent centile effect dcent centile effect dcent centile effect dec1 0.981* 45 0.969* 45 0.725* 35 0.657* 32 0.573* 28 0.534* 26 (0.002) (0.002) (0.003) (0.004) (0.004) (0.005) dec2 0.517* 25 0.507* 25 0.513* 25 0.441* 22 0.359* 18 0.320* 16 (0.002) (0.002) (0.002) (0.004) (0.004) (0.004) dec3 0.373* 18 0.363* 18 0.376* 19 0.297* 15 0.216* 11 0.177* 9 (0.001) (0.002) (0.002) (0.004) (0.004) (0.004) dec4 0.185* 9 0.174* 9 0.172* 9 0.089* 4 0.012* 1

• 0.026*
• 1

(0.001) (0.002) (0.002) (0.004) (0.004) (0.004) dec5 0.117* 6 0.105* 5 0.104* 5 0.018* 1

• 0.073*
• 4
• 0.112*
• 6

(0.001) (0.002) (0.002) (0.003) (0.004) (0.004) dec6 0.080* 4 0.066* 3 0.071* 4

• 0.016*
• 1
• 0.126*
• 6
• 0.167*
• 8

(0.001) (0.002) (0.002) (0.003) (0.004) (0.004) dec7 0.059* 3 0.045* 2 0.053* 3

• 0.034*
• 2
• 0.160*
• 8
• 0.201*
• 10

(0.001) (0.002) (0.002) (0.003) (0.004) (0.004) dec8 0.016* 1 0.000 0.013* 1

• 0.075*
• 4
• 0.214*
• 11
• 0.254*
• 13

(0.001) (0.002) (0.002) (0.003) (0.004) (0.004) dec9

• 0.095*
• 5
• 0.110*
• 5
• 0.063*
• 3
• 0.150*
• 7
• 0.295*
• 15
• 0.335*
• 17

(0.001) (0.002) (0.002) (0.003) (0.004) (0.004) dec10

• 0.250*
• 12
• 0.265*
• 13
• 0.140*
• 7
• 0.224*
• 11
• 0.380*
• 19
• 0.419*
• 21

(0.001) (0.002) (0.002) (0.003) (0.004) (0.004) pichincha 0.007* 0.016* 1 0.017* 1 0.004 0.004 (0.002) (0.002) (0.002) (0.002) (0.002) guayas 0.016* 1 0.024* 1 0.025* 1 0.014* 1 0.013* 1 (0.002) (0.002) (0.002) (0.002) (0.002) coast

• 0.020*
• 1
• 0.015*
• 1
• 0.016*
• 1
• 0.016*
• 1
• 0.017*
• 1

(0.002) (0.002) (0.002) (0.002) (0.002)

Continued on next page

42

(1) (2) (3) (4) (5) (6) dcent centile effect dcent centile effect dcent centile effect dcent centile effect dcent centile effect dcent centile effect (0.002) (0.002) (0.002) (0.002) (0.002) center 0.045* 2 0.032* 2 0.032* 2 0.022* 1 0.022* 1 (0.002) (0.002) (0.002) (0.002) (0.002) south 0.039* 2 0.039* 2 0.041* 2 0.031* 2 0.030* 1 (0.002) (0.002) (0.002) (0.002) (0.002) age19

• 0.075*
• 4
• 0.108*
• 5
• 0.104*
• 5

(0.005) (0.005) (0.005) age20_29 0.084* 4 0.041* 2 0.044* 2 (0.003) (0.003) (0.003) age30_39 0.097* 5 0.061* 3 0.063* 3 (0.003) (0.003) (0.003) age40_49 0.092* 5 0.060* 3 0.061* 3 (0.003) (0.003) (0.003) age50_59 0.086* 4 0.069* 3 0.069* 3 (0.003) (0.003) (0.003) gender 0.022* 1 0.067* 3 (0.001) (0.003) married 0.018* 1 (0.001) education 0.171* 9 (0.001) marriedman 0.025* 1 (0.001) marriedwoman 0.006* (0.002) educman 0.157* 8 (0.001) educwoman 0.221* 11 (0.003) Obs. 1408497 1408497 737 891 737 891 737 891 737 891 F-statistic - full model 54200.9 36331.2 17541.5 13373.3 12751.5 11764.0 R2 0.278 0.279 0.263 0.266 0.284 0.285 Root MSE 0.534 0.533 0.417 0.416 0.411 0.410 * p<0.01

Concluding remarks

• Income mobility at the top of the distribution is low and it

remains stable over the 2004 - 2011 period.

• Top income individuals are more likely to move between the

top 5% and the top 0,1% of the distribution.

• The proportion of individuals who drop to the bottom 95% is

inferior to the proportion of individuals who remains into the top 5% by the final year.

• There is an important degree of mobility in the middle of

income distribution. More than 50% of individuals moved to a higher decile group over the 2008-2011 period.

43

Concluding remarks

• Third, results of regressions analysis suggest

that initial position in the income distribution is closely associated with the probability of upward mobility or downward mobility.

• Moreover, having a high school degree is

associated with moving up in the income distribution by about 10 centiles between 2008 and 2011.

44

Income Mobility in Ecuador: New evidence from individual income tax returns

Liliana CANO University of Toulouse - Lereps September 6th, 2014 INEQUALITY measurement, trends, impacts and policies

45